Executive brief
OpenClaw is an AI assistant platform that includes voice calling capabilities. A security flaw in its webhook verification process allowed attackers to bypass replay protection by slightly modifying the web address of a request. This could allow an attacker to repeatedly trigger the same voice call action, potentially leading to unauthorized operations or service abuse.
Technical details
A replay attack vulnerability exists in OpenClaw's Plivo V2 webhook verification logic within `extensions/voice-call/src/webhook-security.ts`. While the signature validation correctly canonicalized the base URL, the replay key (used to prevent duplicate requests) was derived from the full URL including the query string. Because Plivo V2 signatures only authenticate the base URL and nonce, an attacker could append arbitrary, unsigned query parameters to a valid request to generate a new, unique replay key. This allows the same signed request to be processed multiple times as if it were fresh. The issue is fixed in version 2026.3.23 by ensuring the replay key is derived from the base URL without query parameters.
Affected products
- OpenClaw openclaw < 2026.3.23
Timeline
- 2026-03-24: disclosed
- 2026-03-24: patched
- 2026-03-26: advisory
References
- https://api.github.com/users/smaeljaish771
- https://github.com/smaeljaish771
- https://api.github.com/users/smaeljaish771/gists%7B/gist_id%7D
- https://api.github.com/users/smaeljaish771/repos
- https://avatars.githubusercontent.com/u/266604088?v=4
- https://api.github.com/users/smaeljaish771/events%7B/privacy%7D